How to modify a tf variable using while loop iterator

Question

I would like to slice a tensor and store it in a variable. Slicing with fixed numbers works fine eg: t[0:2]. But slicing a variable with another tensor doesnt work. eg t[t1:t2] Also storing the slice in a tensor works fine but when I try to store it in a tf.Variable i get errors.

import tensorflow as tf
import numpy

i=tf.zeros([2,1],tf.int32)
i2=tf.get_variable('i2_variable',initializer=i) #putting a multidimensional tensor in a variable
i4=tf.ones([10,1],tf.int32)
sess=tf.Session()
sess.run(tf.global_variables_initializer()) #initializing variables
itr=tf.constant(0,tf.int32)

def w_c(i2,itr):
    return tf.less(itr,2)

def w_b(i2,itr):
    i2=i4[(itr*0):((itr*0)+2)] #doesnt work
    #i2=i4[0:2] #works
    #i=i4[(itr*0):((itr*0)+2)] #works with tensor i 
    itr=tf.add(itr,1)
    return[i2,itr]
OP=tf.while_loop(w_c,w_b,[i2,itr])
print(sess.run(OP))

I get following error:

ValueError: Input tensor 'i2_variable/read:0' enters the 
loop with shape (2, 1), but has shape (?, 1) after one iteration. 
To allow the shape to vary across iterations, 
use the `shape_invariants` argument of tf.while_loop to specify a less-specific shape.

Mohan Radhakrishnan · Answer 1 · 2018-07-25T02:24:24.150

1

The code does not throw the error if you specify shape_invariants.

   OP=tf.while_loop(w_c,w_b,[i2,itr],shape_invariants=
                                        [ tf.TensorShape([None, None]), 
                                          itr.get_shape()])

It returns this.

  [array([[1],
   [1]]), 2]

edited Jul 25 '18 at 02:24

answered Jul 18 '18 at 09:22

Mohan Radhakrishnan

3,002
5
28
42

is it mandatory to set shape_invariants even when shape of i2 is same as the value inserted into it. – Aseem Jul 18 '18 at 10:32
Not able to give you an expert opinion. But based on the message `loop with shape (2, 1), but has shape (?, 1) after one iteration.` I relaxed it to be more general. You may also read [this](https://haosdent.gitbooks.io/tensorflow-document/content/api_docs/python/control_flow_ops.html) – Mohan Radhakrishnan Jul 18 '18 at 10:47
I tried tf.assign(i2, i4[(itr*0):((itr*0)+2)] ). It works normally but doesnt work inside a while loop body.. I get error saying Tensor object has no attribute 'assign' . – Aseem Jul 19 '18 at 07:58
Variable i2 is not getting updated coz its not assigned and returned in the while loop. tf.assign returns a tensor so you need to save it into a node and call that node as a while loop variable. 'sess.run([OP,i2])' add this to the end of your code and you will see that i2 value is unchanged. – Aseem Jul 24 '18 at 21:22
I have restored the answer to its original state. You may ask another question. – Mohan Radhakrishnan Jul 25 '18 at 03:31
That's what I am doing (see https://stackoverflow.com/questions/64244431/input-tensor-name-enters-the-loop-with-shape-but-has-shape-unknown-after) but I keep getting this exception.. – Stefan Falk Oct 08 '20 at 09:25

How to modify a tf variable using while loop iterator

1 Answers1